vjepa2 — AI Assistant Context V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning Meta FAIR Mahmoud Assran∗, Adrien Bardes∗, David Fan∗, Quentin Garrido∗, Russell Howes∗, Mojtaba